Moscow foodservice market

Investors are planning to open a foodservice place in Moscow.

Research objective:

prepare a study of the Moscow foodservice market, find interesting features and present the results obtained, which will help in choosing a suitable place for investors.

Research tasks:

Give recommendations on:

• format of the place (investors are considering a coffee shop, restaurant, pizzeria, pub or bar);
• location,
• prices.

Available information:

a dataset with foodservice places in Moscow, compiled on the basis of data from Yandex Maps and Yandex Business services for the summer of 2022. The information posted in the Yandex Business service could have been added by users or found in publicly available sources. It is purely for reference purposes.

The research process:

1. General Information Study.
2. Data preprocessing.
3. Data analysis.
4. Detailed research.
5. General conclusion.
6. Presentation.

General Information Study

# importing the libraries
import pandas as pd
import os
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px
from folium import Map, Choropleth, Marker
from folium.plugins import MarkerCluster

# saving the dataset to the 'data' variable
pth = 'mos_places.csv'

if os.path.exists(pth):
    data = pd.read_csv(pth, sep=',')
else:
    print('Problem with the file')

# saving a copy of the original dataset
df = data.copy()

# the first five rows of the dataset
data.head()

	name	category	address	district	hours	lat	lng	rating	price	avg_bill	middle_avg_bill	middle_coffee_cup	chain	seats	district_eng
0	WoWфли	cafe	Москва, улица Дыбенко, 7/1	Северный административный округ	ежедневно, 10:00–22:00	55.878494	37.478860	5.0	NaN	NaN	NaN	NaN	0	NaN	Northern Administrative District
1	Четыре комнаты	restaurant	Москва, улица Дыбенко, 36, корп. 1	Северный административный округ	ежедневно, 10:00–22:00	55.875801	37.484479	4.5	выше среднего	Средний счёт:1500–1600 ₽	1550.0	NaN	0	4.0	Northern Administrative District
2	Хазри	cafe	Москва, Клязьминская улица, 15	Северный административный округ	пн-чт 11:00–02:00; пт,сб 11:00–05:00; вс 11:00...	55.889146	37.525901	4.6	средние	Средний счёт:от 1000 ₽	1000.0	NaN	0	45.0	Northern Administrative District
3	Dormouse Coffee Shop	coffee shop	Москва, улица Маршала Федоренко, 12	Северный административный округ	ежедневно, 09:00–22:00	55.881608	37.488860	5.0	NaN	Цена чашки капучино:155–185 ₽	NaN	170.0	0	NaN	Northern Administrative District
4	Иль Марко	pizzeria	Москва, Правобережная улица, 1Б	Северный административный округ	ежедневно, 10:00–22:00	55.881166	37.449357	5.0	средние	Средний счёт:400–600 ₽	500.0	NaN	1	148.0	Northern Administrative District

# checking the number of rows in the dataset
print("Total places in the dataset:", data.shape[0])

Total places in the dataset: 8406

# basic information
data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 8406 entries, 0 to 8405
Data columns (total 15 columns):
 #   Column             Non-Null Count  Dtype  
---  ------             --------------  -----  
 0   name               8406 non-null   object 
 1   category           8406 non-null   object 
 2   address            8406 non-null   object 
 3   district           8406 non-null   object 
 4   hours              7870 non-null   object 
 5   lat                8406 non-null   float64
 6   lng                8406 non-null   float64
 7   rating             8406 non-null   float64
 8   price              3315 non-null   object 
 9   avg_bill           3816 non-null   object 
 10  middle_avg_bill    3149 non-null   float64
 11  middle_coffee_cup  535 non-null    float64
 12  chain              8406 non-null   int64  
 13  seats              4795 non-null   float64
 14  district_eng       8406 non-null   object 
dtypes: float64(6), int64(1), object(8)
memory usage: 985.2+ KB

Conclusions

The dataset contains information about 8,406 places. It is clear from the first lines that there are missing values.

There are fourteen columns in the dataset in total:

• name — place's name;
• category — category;
• address — address;
• district — district of the city;
• hours — opening hours;
• lat — latitude of a geographical point;
• lng — longitude of a geographical point;
• rating — place rating according to user ratings in Yandex Maps;
• price — price category;
• avg_bill — the range of the average bill;
• middle_avg_bill — average bill;
• middle_coffee_cup — the cost of a cup of cappuccino;
• chain — chain or not;
• seats — number of seats.

The columns name, category, address, district, hours, price, avg_bill have object type; lat, lng, rating, middle_avg_bill, middle_coffee_cup, seats have float64 type; chain has int64 type.

We can say there is enough data for analysis. We will prepare them for further work.

Data preprocessing

Checking for duplicates

data.duplicated().sum()

0

There are no duplicates.

Processing of abnormal values

# distribution of values in the dataset
data.describe()

	lat	lng	rating	middle_avg_bill	middle_coffee_cup	chain	seats
count	8406.000000	8406.000000	8406.000000	3149.000000	535.000000	8406.000000	4795.000000
mean	55.750109	37.608570	4.229895	958.053668	174.721495	0.381275	108.421689
std	0.069658	0.098597	0.470348	1009.732845	88.951103	0.485729	122.833396
min	55.573942	37.355651	1.000000	0.000000	60.000000	0.000000	0.000000
25%	55.705155	37.538583	4.100000	375.000000	124.500000	0.000000	40.000000
50%	55.753425	37.605246	4.300000	750.000000	169.000000	0.000000	75.000000
75%	55.795041	37.664792	4.400000	1250.000000	225.000000	1.000000	140.000000
max	55.928943	37.874466	5.000000	35000.000000	1568.000000	1.000000	1288.000000

The minimum value of the average bill looks abnormal, the maximum may be explained by the fact that there are restaurants in the dataset focused on banquets. The maximum cost of a cappuccino cup and the number of seats also look abnormal.

Checking the minimum value in the column middle_avg_bill.

# filter out the rows in which the value of middle_avg_bill is 0
data.query('middle_avg_bill == 0')

	name	category	address	district	hours	lat	lng	rating	price	avg_bill	middle_avg_bill	middle_coffee_cup	chain	seats	district_eng
3688	Кофемания	coffee shop	Москва, улица Новый Арбат, 19	Центральный административный округ	ежедневно, круглосуточно	55.752136	37.587784	4.5	высокие	Средний счёт:от 0 ₽	0.0	NaN	1	200.0	Central Administrative District

The owners of "Coffeemania" on Novy Arbat str. did not specify the average bill. Let's look at the median average score in other cafes of this chain in the Central Administrative District of Moscow and replace the value with the median.

# the median average bill in the "Coffeemania" in the Central Administrative District of Moscow
print(data.query('name == "Кофемания" and district_eng == "Central Administrative District"')['middle_avg_bill'].median())

# replace the missing value with the found one
data.loc[3688, 'middle_avg_bill'] = data.query('name == "Кофемания" and district_eng == "Central Administrative District"')['middle_avg_bill'].median()

2000.0

We look at the 95th and 99th percentiles of the cost of a cup of cappuccino.

print(np.percentile(data.query('middle_coffee_cup.notna()', engine="python")['middle_coffee_cup'], [95, 99]))

[275.  309.9]

Coffee shops with the most expensive coffee.

# filter out the places where the cost of a cup of cappuccino is more than 310 rubles
data.query('middle_coffee_cup > 310')

	name	category	address	district	hours	lat	lng	rating	price	avg_bill	middle_avg_bill	middle_coffee_cup	chain	seats	district_eng
1769	Шоколадница	coffee shop	Москва, Бутырская улица, 95	Северный административный округ	пн-чт 07:30–23:00; пт 07:30–00:00; сб круглосу...	55.807331	37.580716	4.1	средние	Цена чашки капучино:300–350 ₽	NaN	325.0	1	61.0	Northern Administrative District
2075	Il Tocco	coffee shop	Москва, Ходынский бульвар, 11А	Северный административный округ	пн-пт 08:00–21:00; сб,вс 09:00–22:00	55.785455	37.530359	4.1	средние	Цена чашки капучино:280–350 ₽	NaN	315.0	0	NaN	Northern Administrative District
2859	Шоколадница	coffee shop	Москва, Большая Семёновская улица, 27, корп. 1	Восточный административный округ	ежедневно, 08:00–23:00	55.782268	37.709022	4.2	средние	Цена чашки капучино:230–2907 ₽	NaN	1568.0	1	48.0	Eastern Administrative District
3718	Кафетериус	coffee shop	Москва, Большая Никитская улица, 35	Центральный административный округ	пн-пт 08:00–22:00; сб,вс 10:00–22:00	55.757292	37.595033	4.3	средние	Цена чашки капучино:279–378 ₽	NaN	328.0	1	30.0	Central Administrative District
5503	Coffee FM	coffee shop	Москва, Авиамоторная улица, 10, корп. 1	Юго-Восточный административный округ	пн-пт 08:00–21:00; сб,вс 09:00–19:00	55.754233	37.715491	4.3	NaN	Цена чашки капучино:250–500 ₽	NaN	375.0	0	190.0	South-Eastern Administrative District
5990	Диемм	coffee shop	Москва, 3-я Фрунзенская улица, 1	Центральный административный округ	ежедневно, 08:30–23:00	55.719000	37.582203	4.3	средние	Цена чашки капучино:250–390 ₽	NaN	320.0	0	30.0	Central Administrative District

Incorrect value in the "Shokoladnitsa": most likely due to a typo in the maximum cost of a cup of coffee. Let's look at the median average score in other places of this chain in the Eastern Administrative District of Moscow and replace the value with the median.

# the median average bill in the "Shokoladnitsa" chain in Eastern Administrative District of Moscow
print(data.query('name == "Шоколадница" and district_eng == "Eastern Administrative District"')['middle_coffee_cup'].median())

# replace the missing value with the found one
data.loc[2859, 'middle_coffee_cup'] = data.query('name == "Шоколадница" and district_eng == "Eastern Administrative District"')['middle_coffee_cup'].median()

256.0

Let's look at the 95th and 99th percentiles of the number of seats and at the places where there are more seats than the 99th percentile.

print(np.percentile(data.query('seats.notna()', engine="python")['seats'], [95, 99]))

[307. 625.]

# filter out establishments where the number of seats is more than 625
data.query('seats > 625')

	name	category	address	district	hours	lat	lng	rating	price	avg_bill	middle_avg_bill	middle_coffee_cup	chain	seats	district_eng
2713	Ваня и Гоги	bar, pub	Москва, Измайловское шоссе, 71, корп. А	Восточный административный округ	ежедневно, 11:00–06:00	55.789842	37.750282	4.2	высокие	Средний счёт:1000–2500 ₽	1750.0	NaN	0	1040.0	Eastern Administrative District
2722	Маргарита	fast food	Москва, Измайловское шоссе, 71, корп. А	Восточный административный округ	ежедневно, 10:00–22:00	55.789527	37.752004	4.3	NaN	NaN	NaN	NaN	1	1040.0	Eastern Administrative District
2770	Шоколадница	coffee shop	Москва, Измайловское шоссе, 71, корп. А	Восточный административный округ	пн-ср 11:00–23:00; чт 11:00–00:00; пт,сб 11:00...	55.789903	37.749822	4.1	NaN	NaN	NaN	NaN	1	1040.0	Eastern Administrative District
2901	Ресторан Тройка	bar, pub	Москва, Измайловское шоссе, 71, корп. 2Б	Восточный административный округ	ежедневно, 11:00–23:00	55.789276	37.747832	3.7	NaN	NaN	NaN	NaN	0	660.0	Eastern Administrative District
2913	Хаус бар	bar, pub	Москва, Измайловское шоссе, 71, корп. 2Б	Восточный административный округ	ежедневно, 07:00–23:00	55.789639	37.747274	3.7	средние	Средний счёт:500–1000 ₽	750.0	NaN	0	660.0	Eastern Administrative District
2966	Матрешка	cafe	Москва, Измайловское шоссе, 71, корп. А	Восточный административный округ	NaN	55.789867	37.749656	4.0	NaN	NaN	NaN	NaN	0	1040.0	Eastern Administrative District
4180	Eataly	bar, pub	Москва, Киевская улица, 2	Западный административный округ	ежедневно, 12:00–23:00	55.743405	37.562535	4.6	NaN	NaN	NaN	NaN	0	920.0	Western Administrative District
4231	РестоБар Argomento	cafeteria	Москва, Кутузовский проспект, 41, стр. 1	Западный административный округ	ежедневно, 12:00–23:00	55.738237	37.531819	4.2	высокие	Средний счёт:2500–5000 ₽	3750.0	NaN	0	1200.0	Western Administrative District
4245	Стейк & Бургер	cafe	Москва, Киевская улица, 2	Западный административный округ	ежедневно, 09:00–21:00	55.742953	37.561872	4.2	NaN	NaN	NaN	NaN	1	920.0	Western Administrative District
5486	Дом	cafe	Москва, улица Юности, 1	Восточный административный округ	NaN	55.736204	37.815500	4.9	NaN	NaN	NaN	NaN	0	760.0	Eastern Administrative District
5655	The Fox Pub	bar, pub	Москва, Мичуринский проспект, 22, корп. 1	Западный административный округ	пн-чт 12:00–00:00; пт,сб 12:00–02:00; вс 12:00...	55.701625	37.504137	4.7	NaN	NaN	NaN	NaN	0	650.0	Western Administrative District
5720	Ресторан китайской кухни Чуаньюй	restaurant	Москва, Мичуринский проспект, 7, корп. 1	Западный административный округ	ежедневно, 11:00–23:00	55.700431	37.512799	4.4	средние	NaN	NaN	NaN	0	650.0	Western Administrative District
5738	Университетское	coffee shop	Москва, Мичуринский проспект, 8, стр. 1	Западный административный округ	ежедневно, 07:00–20:00	55.705561	37.511163	4.3	NaN	NaN	NaN	NaN	0	650.0	Western Administrative District
5758	Шоколадница	coffee shop	Москва, Мичуринский проспект, 22, корп. 1	Западный административный округ	ежедневно, 08:00–23:00	55.701211	37.503986	4.2	средние	Цена чашки капучино:239–274 ₽	NaN	256.0	1	650.0	Western Administrative District
5835	Lyanson’s coffee	cafe	Москва, Мичуринский проспект, 27, корп. 1	Западный административный округ	пн-пт 08:00–20:00	55.697630	37.501684	3.8	NaN	NaN	NaN	NaN	0	650.0	Western Administrative District
5841	For Your Kids	cafe	Москва, Мичуринский проспект, 58, корп. 1	Западный административный округ	NaN	55.691382	37.486793	3.3	NaN	NaN	NaN	NaN	0	650.0	Western Administrative District
6518	DelonixCafe	restaurant	Москва, проспект Вернадского, 94, корп. 1	Западный административный округ	ежедневно, круглосуточно	55.652577	37.475730	4.1	высокие	Средний счёт:1500–2000 ₽	1750.0	NaN	0	1288.0	Western Administrative District
6524	Ян Примус	restaurant	Москва, проспект Вернадского, 121, корп. 1	Западный административный округ	пн-чт 12:00–00:00; пт,сб 12:00–02:00; вс 12:00...	55.657166	37.481519	4.5	выше среднего	Средний счёт:1500 ₽	1500.0	NaN	1	1288.0	Western Administrative District
6548	Vibes cafe	cafe	Москва, улица Миклухо-Маклая, 6	Юго-Западный административный округ	пн-пт 09:00–20:00; сб 09:00–16:00	55.651882	37.499295	4.4	NaN	NaN	NaN	NaN	0	644.0	South-Western Administrative District
6574	Мюнгер	pizzeria	Москва, проспект Вернадского, 97, корп. 1	Западный административный округ	пн-пт 08:00–21:00; сб,вс 10:00–21:00	55.667505	37.491001	4.8	NaN	NaN	NaN	NaN	1	1288.0	Western Administrative District
6641	One Price Coffee	coffee shop	Москва, проспект Вернадского, 84, стр. 1	Западный административный округ	ежедневно, 08:30–20:00	55.665129	37.478635	4.3	NaN	NaN	NaN	NaN	1	1288.0	Western Administrative District
6658	ГудБар	bar, pub	Москва, проспект Вернадского, 97, корп. 1	Западный административный округ	пн-пт 11:00–23:00; сб,вс 13:00–23:00	55.667327	37.490601	4.1	средние	Средний счёт:700 ₽	700.0	NaN	0	1288.0	Western Administrative District
6684	Пивной ресторан	bar, pub	Москва, проспект Вернадского, 121, корп. 1	Западный административный округ	NaN	55.657133	37.481508	4.5	NaN	NaN	NaN	NaN	0	1288.0	Western Administrative District
6690	Японская кухня	restaurant	Москва, проспект Вернадского, 121, корп. 1	Западный административный округ	NaN	55.657255	37.481547	4.4	NaN	NaN	NaN	NaN	1	1288.0	Western Administrative District
6696	Кабул	restaurant	Москва, улица Миклухо-Маклая, 6	Юго-Западный административный округ	пн-пт 07:00–21:00; сб 07:00–18:00	55.652025	37.498719	4.3	NaN	NaN	NaN	NaN	1	644.0	South-Western Administrative District
6771	Точка	cafe	Москва, проспект Вернадского, 84, стр. 1	Западный административный округ	NaN	55.665634	37.477830	4.7	NaN	NaN	NaN	NaN	1	1288.0	Western Administrative District
6807	Loft-cafe академия	cafe	Москва, проспект Вернадского, 84, стр. 1	Западный административный округ	пн-пт 09:00–20:00; сб 09:00–16:00	55.665142	37.478603	3.6	NaN	NaN	NaN	NaN	0	1288.0	Western Administrative District
6808	Яндекс Лавка	restaurant	Москва, проспект Вернадского, 51, стр. 1	Западный административный округ	ежедневно, круглосуточно	55.672580	37.507753	4.0	NaN	NaN	NaN	NaN	1	1288.0	Western Administrative District
6838	Alternative coffee	coffee shop	Москва, проспект Вернадского, 41, стр. 1	Западный административный округ	пн-пт 09:00–21:00; сб,вс 09:00–22:00	55.673128	37.502992	4.3	NaN	NaN	NaN	NaN	0	1288.0	Western Administrative District
7987	Ресторан	restaurant	Москва, улица Маршала Захарова, 6, корп. 1	Южный административный округ	NaN	55.623680	37.704937	4.5	NaN	NaN	NaN	NaN	0	675.0	Southern Administrative District

Judging by the fact that there are groups of places with a very high number of seats that are located at the same address, these are really anomalies. We will leave them in the dataset: for further work with the seats, we will use median values.

Also we have "Yandex Lavka" in the dataset, which is a food delivery service. Let's see how many such names are in the data.

# filter out places with the name "Yandex Lavka" or "Yandex.Lavka" and find their number
data.query('name == "Яндекс Лавка"  or name == "Яндекс.Лавка"').shape[0]

72

Delete them from the dataset.

data = data.query('name not in ("Яндекс Лавка","Яндекс.Лавка")').copy()

Processing of missing values

# count the number of missing values in each column of the dataset
data.isna().sum()

name                    0
category                0
address                 0
district                0
hours                 535
lat                     0
lng                     0
rating                  0
price                5019
avg_bill             4518
middle_avg_bill      5185
middle_coffee_cup    7799
chain                   0
seats                3574
district_eng            0
dtype: int64

Heat map of missing values in percentage ratio.

pd.DataFrame(round(data.isna().mean()*100,)).style.background_gradient('coolwarm')

	0
name	0.000000
category	0.000000
address	0.000000
district	0.000000
hours	6.000000
lat	0.000000
lng	0.000000
rating	0.000000
price	60.000000
avg_bill	54.000000
middle_avg_bill	62.000000
middle_coffee_cup	94.000000
chain	0.000000
seats	43.000000
district_eng	0.000000

More than 50% of the missing values in the price, avg_bill, middle_avg_bill columns: the dataset is based on data from Yandex Maps and Yandex Business services, which means that the owners of places or visitors did not add information about prices and the average bill.

The missing values in the middle_avg_bill column are also due to the fact that data from avg_bill does not get into it if they are specified for coffee shops or bars/pubs.

A very high percentage of omissions (94%) in the middle_coffee_cup column is explained by the fact that this column is filled from the avg_bill column if the value starts with the substring "Price of one cup of cappuccino", i.e. mainly for coffee shops, and there are other categories of places in the dataset.

Fill in the missing values as follows:

• hours — 6% of missing values, we will leave it as it is;
• price — we take information about prices in chain places from other places of the chain that are located in the same administrative district;
• we will leave the column avg_bill with missing values, because for the study we need middle_avg_bill;
• middle_avg_bill — we fill it out based on the assumption that for places of the same category with the same price level and located in the same administrative district, the average bill will be the same.
• middle_coffee_cup — we take information about prices in chain coffee shops from other places of the chain that are located in the same administrative district;
• we will leave the seats column with missing values, because even places of the same chain can have different areas and different numbers of seats.

Fill in the missing values in the price column.

# we sort the chain placws that have a price category specified, and save it to the 'chains' variable
chains = data.query('chain == 1 and price.notna()', engine="python").copy() 

# we collect the price categories in the table, broken down by administrative districts and chains
chains = chains.groupby(['district', 'name']).agg({'price': 'first'})

# the function of replacing the missing price value with the value from the 'chains' table
def find_price_category(row):
    try:
        district = row['district']
        name = row['name']
        return chains.loc[district, name]['price']
    except:
        return None

# apply the function to the price column of the 'data' dataset
data.loc[data['price'].isna(), 'price'] = data.apply(find_price_category, axis=1)

Fill in the missing values in the column middle_avg_bill.

# sorting placas for which the price and middle_avg_bill columns are filled in
places = data.query('price.notna() and middle_avg_bill.notna()', engine="python").copy() 

# collect the price categories and the average bill in the table, by administrative districts and categories of places
places = places.groupby(['district', 'category', 'price']).agg({'middle_avg_bill': 'median'}).round()

# the function of replacing the missing value middle_avg_bill with the value from the 'places' table
def find_middle_avg_bill(row):
    try:
        district = row['district']
        category = row['category']
        price = row['price']
        return places.loc[district, category, price]['middle_avg_bill']
    except:
        return None

# apply the function to the middle_avg_bill column of the 'data' dataset
data.loc[data['middle_avg_bill'].isna(), 'middle_avg_bill'] = data.apply(find_middle_avg_bill, axis=1)

Fill in the missing values in the column middle_coffee_cup.

# sorting the chain coffee shops for which the middle_coffee_cup column is filled
coffee_shops = data.query('chain == 1 and middle_coffee_cup.notna()', engine="python").copy() 

# collect the price categories and the average bill in the table, by administrative districts and categories of places
coffee_shops = coffee_shops.groupby(['district', 'name']).agg({'middle_coffee_cup': 'first'})

# function of replacing the missing value middle_coffee_cup with the value from the coffee_shops table
def find_middle_coffee_cup(row):
    try:
        district = row['district']
        name = row['name']
        return coffee_shops.loc[district, name]['middle_coffee_cup']
    except:
        return None

# apply the function to the middle_coffee_cup column of the 'data' dataset
data.loc[data['middle_coffee_cup'].isna(), 'middle_coffee_cup'] = data.apply(find_middle_coffee_cup, axis=1)

Checking what percentage of passes were processed.

# heat maps for the processed dataset 'data' and the original 'df'
display(pd.DataFrame(round(data.isna().mean()*100,)).style.background_gradient('coolwarm'))
display(pd.DataFrame(round(df.isna().mean()*100,)).style.background_gradient('coolwarm'))

	0
name	0.000000
category	0.000000
address	0.000000
district	0.000000
hours	6.000000
lat	0.000000
lng	0.000000
rating	0.000000
price	54.000000
avg_bill	54.000000
middle_avg_bill	50.000000
middle_coffee_cup	91.000000
chain	0.000000
seats	43.000000
district_eng	0.000000

	0
name	0.000000
category	0.000000
address	0.000000
district	0.000000
hours	6.000000
lat	0.000000
lng	0.000000
rating	0.000000
price	61.000000
avg_bill	55.000000
middle_avg_bill	63.000000
middle_coffee_cup	94.000000
chain	0.000000
seats	43.000000
district_eng	0.000000

The percentage of missing values in the price column decreased from 61 to 54, in middle_avg_bill from 63 to 50, and in middle_coffee_cup from 94 to 91.

Adding columns

We add the street column with street names from the address column to the dataset, the is_24/7 column with the logical value True if the place is open daily and around the clock, and False if not.

# adding a column with street names
data['street'] = data['address'].apply(lambda x: x.split(',')[1].strip())

# adding a column indicating that the place is open daily and around the clock
data['is_24/7'] = data['hours'].apply(lambda x: True if x == "ежедневно, круглосуточно" else False)

Earlier we found out that there are no obvious duplicates in the dataset. Let's check if there are any implicit ones: we will bring the columns with the address and name to the same register, replace ё with е, remove the special characters, then check for duplicates again.

# names and addresses to lowercase
data['address'] = data['address'].str.lower()
data['name'] = data['name'].str.lower()

# replacing ё with е in names and addresses
data['name'] = data['name'].str.replace('ё', 'е')
data['address'] = data['address'].str.replace('ё', 'е')

# duplicates in the name and address columns
data[data[['name', 'address']].duplicated()]

	name	category	address	district	hours	lat	lng	rating	price	avg_bill	middle_avg_bill	middle_coffee_cup	chain	seats	district_eng	street	is_24/7
215	кафе	cafe	москва, парк ангарские пруды	Северный административный округ	ежедневно, 10:00–22:00	55.881438	37.531848	3.2	None	NaN	NaN	NaN	0	NaN	Northern Administrative District	парк Ангарские пруды	False
1511	more poke	restaurant	москва, волоколамское шоссе, 11, стр. 2	Северный административный округ	пн-чт 09:00–18:00; пт,сб 09:00–21:00; вс 09:00...	55.806307	37.497566	4.2	None	NaN	NaN	NaN	1	188.0	Northern Administrative District	Волоколамское шоссе	False
2420	раковарня клешни и хвосты	bar, pub	москва, проспект мира, 118	Северо-Восточный административный округ	пн-чт 12:00–00:00; пт,сб 12:00–01:00; вс 12:00...	55.810677	37.638379	4.4	None	NaN	NaN	NaN	1	150.0	North-Eastern Administrative District	проспект Мира	False
3109	хлеб да выпечка	cafe	москва, ярцевская улица, 19	Западный административный округ	NaN	55.738449	37.410937	4.1	None	NaN	NaN	NaN	0	276.0	Western Administrative District	Ярцевская улица	False

There are 4 implicit duplicates in the dataset. Delete them.

data = data.drop_duplicates(['name', 'address'])

Conclusions

One abnormal value was processed: the average bill of 0.00 in one of the coffee shops of the 'Shokoladnitsa' chain was replaced by the median bill in the coffee shops of this chain in the same administrative district. Removed 4 duplicates.

There are a large number of missing values in the dataset. More than 50% of the missing values in the columns price, avg_bill, middle_avg_bill are explained by the fact that the dataset is based on data from Yandex Maps and Yandex Business services, which means that the owners of places or visitors have not added information about prices and the average bill.

The missing values in the middle_avg_bill column are also due to the fact that data from avg_bill does not get into it if they are specified for coffee shops or bars/pubs.

A very high percentage of missing values (94%) in the middle_coffee_cup column is explained by the fact that this column is filled only for coffee shops.

Fill in the missing values as follows:

• hours, avg_bill, seats as they are;
• for price we take information about prices in chain places from other places of the chain that are located in the same administrative district;
• the middle_avg_bill column was filled with the median value for places of the same category with the same price level and located in the same administrative district;
• the middle_coffee_cup was filled in for chain coffee shops by analogy with other chain places located in the same administrative district.

After processing, the percentage of missing values in the price column decreased from 61 to 54, in middle_avg_bill from 63 to 50, and in middle_coffee_cup from 94 to 91.

Added to the dataset column street with street names, column is_24/7 with the designation that the place is open daily and around the clock.

Data analysis

Number of places by category

Let's see which categories of places and how many of them are represented in the data.

# we group the data by categories and count the number of places in each of them, 
# save them in the 'categories' variable
categories = (
    data.groupby('category') \
    .agg(number=('category', 'count')) \
    .sort_values(by='number', ascending=False)) \
    .reset_index() \

categories

	category	number
0	cafe	2376
1	restaurant	1970
2	coffee shop	1413
3	bar, pub	764
4	pizzeria	633
5	fast food	603
6	cafeteria	315
7	bakery	256

# barplot with the number of places by category

plt.style.use('seaborn')
plt.figure(figsize=(12,7))
plt.title('Number of foodservice places by category',fontsize=15)

g = sns.barplot(data=categories, x='category', y='number')
for p in g.patches:
    g.annotate(round(p.get_height()), xy=(p.get_x() + p.get_width() / 2, p.get_height()), ha = 'center', va = 'center', xytext = (0, 10), textcoords = 'offset points')

plt.xlabel('Categories of places',fontsize=12)
plt.ylabel('Number',fontsize=12);

Cafes and restaurants are represented the most in the dataset — 2,378 and 1,971 places, respectively. The least number of bakeries, 256.

Number of seats in places

# we group the data by categories and calculate the number of seats in each, save it to the 'seats' variable
seats = (
    data.groupby('category') \
    .agg(number=('seats', 'median')) \
    .sort_values(by='number', ascending=False)) \
    .reset_index()

seats

	category	number
0	restaurant	89.0
1	bar, pub	82.0
2	coffee shop	80.0
3	cafeteria	75.5
4	fast food	65.0
5	cafe	60.0
6	pizzeria	55.0
7	bakery	50.0

# barplot with the number of seats by category

plt.figure(figsize=(12,8))
plt.title('Number of seats in foodservice places',fontsize=15)

g = sns.barplot(data=seats, x='category', y='number')
for p in g.patches:
    g.annotate(round(p.get_height()), xy=(p.get_x() + p.get_width() / 2, p.get_height()), ha = 'center', va = 'center', xytext = (0, 10), textcoords = 'offset points')
plt.xlabel('Categories of places',fontsize=12)
plt.ylabel('Number of seats',fontsize=12);

Box-and-whisker diagram.

plt.figure(figsize=(12,8))
plt.title('Number of seats in foodservice places',fontsize=15)

g = sns.boxplot(data=data, x='category', y='seats')
plt.ylim(-10, 1000)
plt.xlabel('Categories of places',fontsize=12)
plt.ylabel('Number of seats',fontsize=12);

Restaurants, bars/pubs and coffee shops are in the lead in terms of the number of seats: the median number of seats is 90, 82 and 80, respectively. On the boxplot, we see outliers: in each category there are places, the number of seats in which significantly exceeds the median.

The ratio of chain and non-chain places

# we group the data according to the chain/non-chain place criterion, calculate the amount, 
# save it to the 'is_chains' variable
is_chains = data.groupby('chain').agg(number=('chain', 'count')).reset_index()
is_chains.loc[0, 'chain'] = "Non-chain"
is_chains.loc[1, 'chain'] = "Chain"
is_chains

	chain	number
0	Non-chain	5199
1	Chain	3131

# pie chart with the ratio of chain and non-chain places

number = is_chains['number']
lables = is_chains['chain']
plt.pie(number, labels=lables, autopct="%1.1f%%", normalize=True)
plt.title("The ratio of chain and non-chain places")
plt.show();

There are 63.2% of non—chain places in the dataset, 36.8% of chain places.

Which categories of places are more often chains?

# we group the data by category and format of the place, calculate the amount, 
# save it to the 'chains_categories' variable

chains_categories = data.groupby(['category', 'chain']).agg(number=('chain', 'count')).reset_index()
chains_categories.loc[chains_categories['chain'] == 0, 'chain'] = 'no'
chains_categories.loc[chains_categories['chain'] == 1, 'chain'] = 'yes'
for index, row in chains_categories.iterrows():
    chains_categories.loc[chains_categories['category'] == row[0],'total'] = data.query('category == @row[0]')['category'].count()
chains_categories['percent'] = round(chains_categories['number'] / chains_categories['total'] * 100, 1)
chains_categories

	category	chain	number	total	percent
0	bakery	no	99	256.0	38.7
1	bakery	yes	157	256.0	61.3
2	bar, pub	no	596	764.0	78.0
3	bar, pub	yes	168	764.0	22.0
4	cafe	no	1597	2376.0	67.2
5	cafe	yes	779	2376.0	32.8
6	cafeteria	no	227	315.0	72.1
7	cafeteria	yes	88	315.0	27.9
8	coffee shop	no	693	1413.0	49.0
9	coffee shop	yes	720	1413.0	51.0
10	fast food	no	371	603.0	61.5
11	fast food	yes	232	603.0	38.5
12	pizzeria	no	303	633.0	47.9
13	pizzeria	yes	330	633.0	52.1
14	restaurant	no	1313	1970.0	66.6
15	restaurant	yes	657	1970.0	33.4

# graph with the share of chain and non-chain places in each category

fig = px.bar(chains_categories, 
             x="percent",
             y="category", 
             color="chain", 
             orientation='h',  width=1000, height=500, 
             color_discrete_sequence=px.colors.qualitative.T10,
             text = chains_categories['percent'].map("{:,}%".format),
             #.map("{:,}%".format),
             labels={"percent": "Percentage of places",
                     "category": "Categories",
                    "chain": "Chain"},
             title="Percentage of chain and non-chain places by category")\
            .update_yaxes(categoryorder="max ascending")
fig.show()

As we can see on the graph, there are three categories of places that are more often chains: coffee shops, pizzerias and bakeries.

Top 15 popular foodservice chains in Moscow

We will consider as a chain those places for which the value 1 is indicated in the chain column, which have the same name and the same category. Popularity will be assessed by the number of places in the chain.

# we sort chain places, group them by name, count the number of places
# and the number of categories so that non-chain places with the same names do not fall into the group
# save it to the 'top_15' variable 

top_15 = (data.query('chain == 1')\
          .groupby('name') \
          .agg(number=('name', 'count'), category=('category', 'first'), count_category=('category', 'nunique')) \
          .sort_values(by=['count_category', 'number'], ascending=[True, False]) \
          .reset_index() \
          .head(15))

top_15

	name	number	category	count_category
0	домино'с пицца	76	pizzeria	1
1	додо пицца	74	pizzeria	1
2	one price coffee	71	coffee shop	1
3	cofix	65	coffee shop	1
4	кофепорт	42	coffee shop	1
5	кулинарная лавка братьев караваевых	39	cafe	1
6	drive café	24	cafe	1
7	cinnabon	20	coffee shop	1
8	штолле	19	bakery	1
9	арамье	18	bakery	1
10	vasilchukí chaihona №1	17	restaurant	1
11	сушистор	16	restaurant	1
12	моремания	15	restaurant	1
13	pizza express 24	14	pizzeria	1
14	бургер кинг	14	restaurant	1

We check how the top 15 is distributed by category.

# group the top 15 by category, count the number

top_15.groupby('category').agg(number=('category', 'count')).sort_values(by='number', ascending=False).reset_index()

	category	number
0	coffee shop	4
1	restaurant	4
2	pizzeria	3
3	bakery	2
4	cafe	2

# barplot with the name of the chains, the number of places and the rating

plt.figure(figsize=(14,7))
plt.title('Top 15 popular chains',fontsize=15)

g = sns.barplot(data=top_15, x='number', y='name', orient='h')

for p in g.patches:
    width, height = p.get_width(), p.get_height()
    x, y = p.get_xy() 
    g.text(x+width - 2, 
           y+height/2,
           '{:.0f}'.format(width), 
           horizontalalignment='left', 
           verticalalignment='center')

plt.xlabel('Number of places',fontsize=12)
plt.ylabel('Names',fontsize=12);

Domino's Pizza and Dodo Pizza chains are leading among the popular chains. The top 15 chains do not include pubs/bars, fast food places, canteens.

Let's take a look at the total number of places from the top 15 and the number of places of each category by district.

# save the names of places from the top 15 chains to the 'top_names' list
top_names = list(top_15['name'].unique())

# filter out places from the top 15 chains and save them to the 'top_15_district' variable
top_15 = data.query('name in @top_names')

# we group the top 15 chains places by district and category, calculate the number
top_15_district = top_15.groupby(['district_eng', 'category']).agg(number=('name', 'count')).reset_index()
top_15_district['district_eng'] = top_15_district['district_eng'].apply(lambda x: x.split(' ')[0])

# interactive chart with the distribution of the top 15 chains places by districts of Moscow

fig = px.bar(top_15_district, 
             x="number",
             y="district_eng", 
             color="category", 
             orientation='h',  width=1000, height=500, 
             color_discrete_sequence=px.colors.qualitative.T10,
             labels={
                     "district_eng": "Moscow District",
                     "number": "Number of places",
                     "category": "Categories"
                 },
             title="Distribution of the top 15 chains categories by districts of Moscow")\
            .update_yaxes(categoryorder="total ascending")
fig.show()

It is expected that most of the places from the top 15 are in the Central Administrative District of Moscow. Least of all, 33, in the North-West.

Places in Moscow districts

Which administrative divisions of Moscow are present in the dataset?

# list of administrative districts of Moscow from the dataset
for x in list(data['district_eng'].unique()):
    print(x)

Northern Administrative District
North-Eastern Administrative District
North-Western Administrative District
Western Administrative District
Central Administrative District
Eastern Administrative District
South-Eastern Administrative District
Southern Administrative District
South-Western Administrative District

In total, 9 administrative districts of Moscow are represented in the dataset. Let's look at the total number of places and the number of places of each category in these districts.

# we collect in the 'districts' pivot table the number of places of each category, 
# divided by administrative districts

districts = data.pivot_table(index='district_eng', columns = 'category', values='name', aggfunc='count').reset_index()
districts['total'] = districts.sum(axis=1, numeric_only=True)
districts['district_eng'] = districts['district_eng'].apply(lambda x: x.split(' ')[0].strip())
districts = districts.sort_values(by='total')
districts;

# for the graph, we group the number of places of each category by administrative districts 
# in the 'districts2' variable 

districts2 = data.groupby(['district_eng', 'category']).agg(number=('category', 'count')).reset_index()
districts2['district_eng'] = districts2['district_eng'].apply(lambda x: x.split(' ')[0])

# interactive chart with the distribution of places of all categories by districts of Moscow

fig = px.bar(districts2, 
             x="number",
             y="district_eng", 
             color="category", 
             orientation='h',  width=1000, height=500, 
             color_discrete_sequence=px.colors.qualitative.T10,
             labels={
                     "district_eng": "Moscow District",
                     "number": "Number of places",
                     "category": "Categories"
                 },
             title="Distribution of categories of all Moscow foodservice places by districts")\
            .update_yaxes(categoryorder="total ascending")
fig.show()

The Central Administrative District of Moscow is also the leader in the total number of places.

Distribution of average ratings by categories of places

Let's look at the boxplot.

plt.figure(figsize=(12,8))
plt.title('Ratings by category of places',fontsize=15)

g = sns.boxplot(data=data, x='category', y='rating')
#plt.ylim(-10, 1000)
plt.xlabel('Categories',fontsize=12)
plt.ylabel('Rating',fontsize=12);

Almost all categories of places have outliers towards low ratings.

# we group the data by categories and calculate the average rating of each, save it to the 'categor_rating' variable
categ_rating = (data.groupby('category') \
                .agg(mean_rating=('rating', 'mean')) \
                .reset_index() \
                .sort_values(by='mean_rating', ascending=False))
categ_rating

	category	mean_rating
1	bar, pub	4.387696
7	restaurant	4.306294
6	pizzeria	4.301264
4	coffee shop	4.277282
0	bakery	4.268359
3	cafeteria	4.211429
2	cafe	4.124285
5	fast food	4.050249

# distribution of average ratings by categories of places

plt.figure(figsize=(12,6))
plt.title('Distribution of average ratings by categories of places',fontsize=15)

g = sns.barplot(data=categ_rating, x='category', y='mean_rating')
for p in g.patches:
    g.annotate('{:.2f}'.format(p.get_height()), xy=(p.get_x() + p.get_width() / 2, p.get_height()), ha = 'center', va = 'center', xytext = (0, 10), textcoords = 'offset points')
plt.xlabel('Categories',fontsize=12)
plt.ylabel('Average rating',fontsize=12)
g.set_ylim(0, 5);

The average rating of all categories exceeds 4 points. Bars/pubs have a maximum rating of 4.39, fast food places have a minimum rating of 4.05.

Average rating of each district places

We calculate the average rating of places in each administrative district of Moscow.

# we group the data by districts and calculate the average rating, save it in the 'district_rating' variable

district_rating = data.groupby('district', as_index=False)['rating'].agg('mean')
district_rating.sort_values(by='rating', ascending=False)

	district	rating
5	Центральный административный округ	4.378543
2	Северный административный округ	4.241275
4	Северо-Западный административный округ	4.209606
8	Южный административный округ	4.187614
1	Западный административный округ	4.187262
0	Восточный административный округ	4.179389
7	Юго-Западный административный округ	4.177937
3	Северо-Восточный административный округ	4.150396
6	Юго-Восточный административный округ	4.106232

We build a choropleth map with an average rating of places in each district.

# uploading a JSON file with the borders of Moscow districts
state_geo = 'admin_level_geomap.geojson'

# moscow_lat - latitude of the center of Moscow, moscow_lng - longitude of the center of Moscow
moscow_lat, moscow_lng = 55.751244, 37.618423

# creating a map of Moscow
m = Map(location=[moscow_lat, moscow_lng], zoom_start=10)

# creating a choropleth map using the Choropleth function and adding it to the map
Choropleth(
    geo_data=state_geo,
    data=district_rating,
    columns=['district', 'rating'],
    key_on='feature.name',
    fill_color='RdPu',
    fill_opacity=0.8,
    legend_name='Average rating of foodservice places by administrative districts of Moscow',
).add_to(m)

# displaying the map
m

Make this Notebook Trusted to load map: File -> Trust Notebook

Places in the Central Administrative District of Moscow have the highest average rating (4.38), the minimum in the South-Eastern (4.1).

Clusters of foodservice places

All places of the dataset displayed on the map using clusters.

# moscow_lat - latitude of the center of Moscow, moscow_lng - longitude of the center of Moscow
moscow_lat, moscow_lng = 55.751244, 37.618423

# creating Moscow map
m = Map(location=[moscow_lat, moscow_lng], zoom_start=10)

# creating an empty cluster, adding it to the map
marker_cluster = MarkerCluster().add_to(m)

# a function that takes a dataframe row,
# creates a marker at the current point and adds it to the 'marker_cluster' cluster
def create_clusters(row):
    Marker(
        [row['lat'], row['lng']],
        popup=f"{row['name']} {row['rating']}",
    ).add_to(marker_cluster)

# applying the create_clusters() function to each row of the dataframe
data.apply(create_clusters, axis=1)

# displaying the map
m

Make this Notebook Trusted to load map: File -> Trust Notebook

There are noticeably more places in the center, north and west of Moscow than in the south and east.

Top 15 streets by number of foodservice places

# we group places by streets and count the number, save the first 15 in the 'top_streets' variable

top_streets = (data.groupby('street') \
               .agg(number=('name', 'count')) \
               .sort_values(by='number', ascending=False) \
               .reset_index() \
               .head(15))
top_streets

	street	number
0	проспект Мира	183
1	Профсоюзная улица	121
2	проспект Вернадского	106
3	Ленинский проспект	106
4	Ленинградский проспект	95
5	Дмитровское шоссе	87
6	Каширское шоссе	77
7	Варшавское шоссе	74
8	Ленинградское шоссе	70
9	МКАД	65
10	Люблинская улица	60
11	улица Вавилова	55
12	Кутузовский проспект	54
13	улица Миклухо-Маклая	48
14	Пятницкая улица	48

Mira Avenue is the leader in terms of the number of places in Moscow, 184 places are located on this street.

Let's plot the distribution of the number of places and their categories by the top 15 streets.

# creating a 'streets' list  with the names of the top 15 streets
streets = list(top_streets['street'])

# collecting places in a table by top-15 streets and categories, calculating the number, 
# saving it to the 'top_streets_categorie' variable 
top_streets_categories = data.query('street in @streets').pivot_table(index='street', columns = 'category', values='name', aggfunc='count').reset_index()
top_streets_categories['total'] = top_streets_categories.sum(axis=1, numeric_only=True)
top_streets_categories = top_streets_categories.fillna(0)
top_streets_categories.loc[:, 'bar, pub': 'total'] = top_streets_categories.loc[:, 'bar, pub': 'total'].astype('int')
top_streets_categories = top_streets_categories.sort_values(by='total')
top_streets_categories

category	street	bakery	bar, pub	cafe	cafeteria	coffee shop	fast food	pizzeria	restaurant	total
10	Пятницкая улица	3.0	9	7	0	6	2	3	18	48
14	улица Миклухо-Маклая	0.0	3	21	0	4	4	2	14	48
3	Кутузовский проспект	1.0	2	14	3	13	2	3	16	54
13	улица Вавилова	2.0	2	15	0	10	11	3	12	55
7	Люблинская улица	0.0	5	26	2	11	5	1	10	60
8	МКАД	0.0	1	45	1	4	9	0	5	65
5	Ленинградское шоссе	2.0	5	13	3	13	5	3	26	70
0	Варшавское шоссе	0.0	6	18	7	14	7	4	18	74
2	Каширское шоссе	0.0	2	20	5	16	10	5	19	77
1	Дмитровское шоссе	2.0	6	23	4	11	10	8	23	87
4	Ленинградский проспект	4.0	15	12	3	25	2	9	25	95
6	Ленинский проспект	3.0	10	26	5	23	2	5	32	106
11	проспект Вернадского	1.0	7	25	2	16	12	12	31	106
9	Профсоюзная улица	4.0	6	35	3	18	15	15	25	121
12	проспект Мира	4.0	11	53	2	36	21	11	45	183

# for the graph, we group by top-15 streets and categories, calculate the number, save it to the 'top_streets_categories2' variable 
top_streets_categories2 = data.query('street in @streets').groupby(['street', 'category']).agg(number=('name', 'count')).reset_index()

# interactive chart with the distribution of places of all categories in the top 15 streets of Moscow

fig = px.bar(top_streets_categories2, 
             x="number",
             y="street", 
             color="category", 
             orientation='h',  width=1000, height=500, 
             color_discrete_sequence=px.colors.qualitative.T10,
             labels={
                     "street": "Street",
                     "number": "Number of places",
                     "category": "Categories"
                 },
             title="Distribution of categories of places by top-15 streets") \
            .update_yaxes(categoryorder="total ascending")
fig.show()

The most numerous categories of places on the top 15 streets are cafes, coffee shops and restaurants.

Streets with one foodservice place

We will find streets where there is only one foodservice place, and see what kind of places they are.

# in the 'streets_with_one' variable, data grouped by streets and number of places
# we filter the streets with one place and display the number of rows of this dataset
streets_with_one = (data.groupby('street') \
               .agg(number=('name', 'count'), category=('category', 'first')) \
               .sort_values(by='number', ascending=False) \
               .query('number == 1') \
               .reset_index())
streets_with_one.shape[0]

459

# grouping streets with one place from the 'streets_with_one' variable by categories
# counting the number of places in each category

streets_with_one.groupby('category', as_index=False)['street'].agg('count').sort_values(by='street', ascending=False)

	category	street
2	cafe	161
7	restaurant	89
4	coffee shop	85
1	bar, pub	39
3	cafeteria	36
5	fast food	23
6	pizzeria	18
0	bakery	8

There are a total of 458 streets in Moscow, where there is only one foodservice place, most often it will be a cafe.

How does distance from the center affect prices in foodservice places?

We calculate the median of the column with the value of the average check middle_avg_bill for each district and use this value as a price indicator of the area.

# we group the data by administrative district, calculate the median average bill, 
# save it to the 'district_avg_bill' variable 

district_avg_bill = data.groupby('district', as_index=False)['middle_avg_bill'].agg('median')
district_avg_bill.sort_values(by='middle_avg_bill', ascending=False)

	district	middle_avg_bill
5	Центральный административный округ	950.0
4	Северо-Западный административный округ	675.0
1	Западный административный округ	650.0
8	Южный административный округ	550.0
0	Восточный административный округ	525.0
7	Юго-Западный административный округ	525.0
2	Северный административный округ	500.0
3	Северо-Восточный административный округ	500.0
6	Юго-Восточный административный округ	425.0

Building a choroplet with the values obtained for each district.

# moscow_lat - latitude of the center of Moscow, moscow_lng - longitude of the center of Moscow
moscow_lat, moscow_lng = 55.751244, 37.618423

# creating Moscow map
m = Map(location=[moscow_lat, moscow_lng], zoom_start=10)

# creating a choropleth map using the Choropleth function and adding it to the map
Choropleth(
    geo_data=state_geo,
    data=district_avg_bill,
    columns=['district', 'middle_avg_bill'],
    key_on='feature.name',
    fill_color='RdPu',
    fill_opacity=0.8,
    legend_name='Median average bill in places by administrative districts of Moscow',
).add_to(m)

# displaying the map
m

Make this Notebook Trusted to load map: File -> Trust Notebook

The median average bill in the Central Administrative District (950 rubles) is significantly higher than in other districts. The cheapest places are located in the South-Eastern Administrative District (the average bill is 425 rubles).

Opening hours

We study the opening hours of places and their dependence on the location and category of the places.

# group the data by category and opening hours, count the number of places

is_day_night = data.groupby(['category','is_24/7']).agg(number=('is_24/7', 'count')).reset_index()
is_day_night = is_day_night.rename(columns={'is_24/7': 'all_around'}) 
is_day_night.loc[is_day_night['all_around'] == False, 'all_around'] = 'no'
is_day_night.loc[is_day_night['all_around'] == True, 'all_around'] = 'yes'

# we count the total number of places in each category 
# and the percentage of places with round-the-clock and opening hours
for index, row in is_day_night.iterrows():
    is_day_night.loc[is_day_night['category'] == row[0],'total'] = data.query('category == @row[0]')['category'].count()
is_day_night['percent'] = round(is_day_night['number'] / is_day_night['total'] * 100, 1)
is_day_night

	category	all_around	number	total	percent
0	bakery	no	232	256.0	90.6
1	bakery	yes	24	256.0	9.4
2	bar, pub	no	712	764.0	93.2
3	bar, pub	yes	52	764.0	6.8
4	cafe	no	2109	2376.0	88.8
5	cafe	yes	267	2376.0	11.2
6	cafeteria	no	303	315.0	96.2
7	cafeteria	yes	12	315.0	3.8
8	coffee shop	no	1354	1413.0	95.8
9	coffee shop	yes	59	1413.0	4.2
10	fast food	no	453	603.0	75.1
11	fast food	yes	150	603.0	24.9
12	pizzeria	no	602	633.0	95.1
13	pizzeria	yes	31	633.0	4.9
14	restaurant	no	1843	1970.0	93.6
15	restaurant	yes	127	1970.0	6.4

# a graph with the share of round-the-clock and regular places in each category

fig = px.bar(is_day_night, 
             x="percent",
             y="category", 
             color="all_around", 
             orientation='h',  width=1000, height=500, 
             color_discrete_sequence=px.colors.qualitative.T10,
             text = is_day_night['percent'].map("{:,}%".format),
             labels={"percent": "Percentage of places",
                     "category": "Categories",
                    "all_around": "24/7"},
             title="Percentage of round-the-clock and regular places by category")\
            .update_yaxes(categoryorder="min ascending")
fig.show()

It is expected that the highest percentage of round-the-clock places are in the "fast food" category, canteens have the lowest percentage.

Let's see in which districts of Moscow there are more round-the-clock places.

# group the data by district and count the number of round-the-clock places, save it to the 'district_day_night' variable 
district_day_night = data.groupby('district', as_index=False)['is_24/7'].agg('sum')
district_day_night

	district	is_24/7
0	Восточный административный округ	95
1	Западный административный округ	71
2	Северный административный округ	71
3	Северо-Восточный административный округ	74
4	Северо-Западный административный округ	43
5	Центральный административный округ	131
6	Юго-Восточный административный округ	92
7	Юго-Западный административный округ	72
8	Южный административный округ	73

# moscow_lat - latitude of the center of Moscow, moscow_lng - longitude of the center of Moscow
moscow_lat, moscow_lng = 55.751244, 37.618423

# creating Moscow map
m = Map(location=[moscow_lat, moscow_lng], zoom_start=10)

# creating a choropleth map using the Choropleth function and adding it to the map
Choropleth(
    geo_data=state_geo,
    data=district_day_night,
    columns=['district', 'is_24/7'],
    key_on='feature.name',
    fill_color='RdPu',
    fill_opacity=0.8,
    legend_name='The number of round-the-clock places in the administrative districts of Moscow',
).add_to(m)

# displaying the map
m

Make this Notebook Trusted to load map: File -> Trust Notebook

The maximum number of round-the-clock places in the Central Administrative District, on the choroplete it is also visible that there are more round-the-clock places in the east of Moscow than in the west.

Characteristic of places with poor ratings

For our study, we will consider a rating below 4 points to be bad. Let's see what percentage of places with a low rating.

# we filter places with a rating below 4 and calculate their percentage

print("Percentage of places with a low rating:", round(data.query('rating < 4')['name'].count() / data['name'].count() * 100))

Percentage of places with a low rating: 14

We look at how low-rated places are distributed by category.

# we group places with a low rating by categories and count their number, save them in 'low_rating_places'
low_rating_places = data.query('rating < 4').groupby('category', as_index=False).agg(low_rating_number=('name', 'count'))

# we group establishments by categories and count their number, save them in 'all_places'
all_places = data.groupby('category', as_index=False).agg(all_places=('name', 'count'))

# we merge two tables, calculate the share of places with a low rating in each category
low_raiting_percent = low_rating_places.merge(all_places, on='category')
low_raiting_percent['share'] = round(low_raiting_percent['low_rating_number'] / low_raiting_percent['all_places'] * 100)
low_raiting_percent

	category	low_rating_number	all_places	share
0	bakery	30	256	12.0
1	bar, pub	40	764	5.0
2	cafe	542	2376	23.0
3	cafeteria	44	315	14.0
4	coffee shop	138	1413	10.0
5	fast food	162	603	27.0
6	pizzeria	35	633	6.0
7	restaurant	153	1970	8.0

Fast food places (27%) and cafes (23%) have the highest share of places with a rating below 4 points.

Conclusions

The following can be said about foodservice places in Moscow:

• Cafes and restaurants are represented the most in the dataset — 2,378 and 1,971 places, respectively. The least number have bakeries (256).

• Restaurants, bars/pubs and coffee shops are leading in terms of the number of seats: the median number of seats is 90, 82 and 80, respectively.

• Non—chain places in the dataset 63.2%, chain - 36.8%. More often they are chain coffee shops, pizzerias and bakeries.

• Domino's Pizza and Dodo Pizza pizza chains are leading among the popular chains. The top 15 chains do not include pubs/bars, fast food places, canteens.

• Most of the places from the top 15 chains are in the Central Administrative District of Moscow (11). Least of all — one — in the East. The Central Administrative District of Moscow is also the leader in the total number of places — 2,237.

• The average rating of all categories of places exceeds 4 points. The maximum rating of bars/pubs is 4.39, fast food places have the minimum rating (4.05). The maximum average rating of places is in the Central Administrative District (4.38), the minimum in the South—East (4.1).

• Mira Avenue is the leader in terms of the number of places in Moscow, 184 places are located on this street. The most numerous categories of places on the top 15 streets: cafes, coffee shops and restaurants. There are 458 streets where there is only one foodservice place in Moscow. Most often the only place will be a cafe.

• The median average bill is in the Central Administrative District (950 rubles) is significantly higher than in other districts. The cheapest places are located in the South-Eastern Administrative District (the average bill is 425 rubles.)

• The highest percentage of round-the-clock places is in the fast food category, the lowest in canteens. The maximum number of round-the-clock places in the Central Administrative District (131), there are more round-the-clock places in the east of Moscow than in the west.

• Fast food places (27%) and cafes (23%) have the highest share of places with a rating below 4 points.

We recommend paying attention to such categories of places as coffee shops and pizzerias: there are fewer of these places than cafes and restaurants – the competition is lower. On the other hand, they are not as limited in format as bars and pubs, which are not suitable, for example, for family holidays.

The Central Administrative District of Moscow is the most saturated with foodservice place – 2,237 establishments. In the city center, the audience is much higher than just the number of residents, due to offices, cultural institutions, etc. To open a new place, we recommend considering the Northern, Northeastern and Southern districts: there is less competition compared to the Central Administrative District, but at the same time there are enough places in these areas, so there is an audience.

As for the average bill, we recommend focusing on the median average bill of places depending on the district. In the Northern, North-Eastern and Southern districts, the median bill is 650 rubles, 500 rubles and 500 rubles, respectively.

Detailed research: opening a coffee shop

Investors are planning to open a coffee shop in Moscow. Let's try to give a recommendation for opening a new place.

Number and location of coffee shops

Let's see how many coffee shops there are in the dataset, in which districts of Moscow there are most of them, what are the features of their location.

print("Total coffee shops in the dataset:", data.query('category == "coffee shop"').shape[0])

Total coffee shops in the dataset: 1413

# we group the filtered coffee shops by district and count the number
district_coffee_shops = data.query('category == "coffee shop"').groupby('district_eng', as_index=False).agg(number=('name', 'count'))
district_coffee_shops['district_eng'] = district_coffee_shops['district_eng'].apply(lambda x: x.split(' ')[0])

# interactive graph of coffee shops distribution by Moscow districts

fig = px.bar(district_coffee_shops, 
             x="number",
             y="district_eng",  
             orientation='h',  width=1000, height=500, 
             color_discrete_sequence=px.colors.qualitative.T10,
             labels={
                     "district_eng": "Moscow district",
                     "number": "Number of coffee shops"
                 },
             title="Distribution of Moscow coffee shops by districts")\
            .update_yaxes(categoryorder="total ascending")
fig.show()

Сlusters of coffee shops on the map of Moscow.

# moscow_lat - latitude of the center of Moscow, moscow_lng - longitude of the center of Moscow
moscow_lat, moscow_lng = 55.751244, 37.618423

# creating Moscow map
m = Map(location=[moscow_lat, moscow_lng], zoom_start=10)

# creating an empty cluster, adding it to the map
marker_cluster = MarkerCluster().add_to(m)

# applying the create_clusters() function to each row of the dataframe with the 'coffee shop' category
data.query('category == "coffee shop"').apply(create_clusters, axis=1)

# displaying the map
m

Make this Notebook Trusted to load map: File -> Trust Notebook

As we know from the study of the distribution of places by districts of Moscow, in the Southern Administrative District, the number of places of which is comparable to the Northern and North-Eastern District, there are relatively fewer coffee shops. We can see this on the map as well.

From the point of view of the share of coffee shops among all the places of the district, one can consider the Southern Administrative District of Moscow to open a coffee shop.

24-hour coffee shops

Previously, we studied the opening hours of foodservice places and know that only 4.2% of all coffee shops in Moscow work 24/7.

Let's see how many round-the-clock coffee shops there are and how they are distributed by districts.

# we group coffee shops by district and count the number of round-the-clock ones, 
# save them to the 'district_is_24_7' variable 

district_is_24_7 = data.query('category == "coffee shop"').groupby('district_eng').agg(number=('is_24/7', 'sum')).sort_values(by='number', ascending=False).reset_index()
district_is_24_7 = district_is_24_7.rename(columns={'district_eng': 'District', 'number': 'Number'})
district_is_24_7

	District	Number
0	Central Administrative District	26
1	Western Administrative District	9
2	South-Western Administrative District	7
3	Eastern Administrative District	5
4	Northern Administrative District	5
5	North-Eastern Administrative District	3
6	North-Western Administrative District	2
7	South-Eastern Administrative District	1
8	Southern Administrative District	1

There is only one round-the-clock coffee shop in the South-Eastern and Southern districts of Moscow. The largest number of 24—hour coffee shops is in the city center (26).

Coffee shop ratings

Previously, we studied the distribution of average ratings by categories of foodservice places and we know that the average rating of coffee shops is 4.28.

Let's see how the ratings of coffee shops are distributed by districts.

districts_coffee_shops_rating = data.query('category == "coffee shop"').groupby('district_eng').agg(mean_rating=('rating', 'mean')).sort_values(by='mean_rating').reset_index()
districts_coffee_shops_rating['mean_rating'] = districts_coffee_shops_rating['mean_rating'].round(2)
districts_coffee_shops_rating

	district_eng	mean_rating
0	Western Administrative District	4.20
1	North-Eastern Administrative District	4.22
2	South-Eastern Administrative District	4.23
3	Southern Administrative District	4.23
4	Eastern Administrative District	4.28
5	South-Western Administrative District	4.28
6	Northern Administrative District	4.29
7	North-Western Administrative District	4.33
8	Central Administrative District	4.34

In Southern District, the average rating of coffee shops is slightly lower than in Moscow as a whole.

# moscow_lat - latitude of the center of Moscow, moscow_lng - longitude of the center of Moscow
moscow_lat, moscow_lng = 55.751244, 37.618423

# creating Moscow map
m = Map(location=[moscow_lat, moscow_lng], zoom_start=10)

# creating a choropleth map using the Choropleth function and adding it to the map
Choropleth(
    geo_data=state_geo,
    data=districts_coffee_shops_rating,
    columns=['district', 'mean_rating'],
    key_on='feature.name',
    fill_color='RdPu',
    fill_opacity=0.8,
    legend_name='Average rating of coffee shops by administrative districts of Moscow',
).add_to(m)

# displaying the map
m

Make this Notebook Trusted to load map: File -> Trust Notebook

The coffee shops in the Central district and the North-Western district have the highest average rating.

The cost of a cup of cappuccino

We analyze what the cost of a cup of cappuccino is worth focusing on when opening a coffee shop. We look at the average price of a cup of cappuccino in coffee shops in Moscow districts.

# we group coffee shops by administrative districts and calculate the average cost of a cup of cappuccino
districts_coffee_shops_cup_price = data.query('category == "coffee shop"').groupby('district_eng').agg(middle_coffee_cup=('middle_coffee_cup', 'mean')).sort_values(by='middle_coffee_cup', ascending=False).reset_index()
districts_coffee_shops_cup_price['middle_coffee_cup'] = districts_coffee_shops_cup_price['middle_coffee_cup'].round()
districts_coffee_shops_cup_price

	district_eng	middle_coffee_cup
0	Central Administrative District	187.0
1	South-Western Administrative District	187.0
2	Western Administrative District	170.0
3	North-Eastern Administrative District	168.0
4	Southern Administrative District	165.0
5	North-Western Administrative District	164.0
6	Northern Administrative District	158.0
7	Eastern Administrative District	147.0
8	South-Eastern Administrative District	137.0

# moscow_lat - latitude of the center of Moscow, moscow_lng - longitude of the center of Moscow
moscow_lat, moscow_lng = 55.751244, 37.618423

# creating Moscow map
m = Map(location=[moscow_lat, moscow_lng], zoom_start=10)

# creating a choropleth map using the Choropleth function and adding it to the map
Choropleth(
    geo_data=state_geo,
    data=districts_coffee_shops_cup_price,
    columns=['district', 'middle_coffee_cup'],
    key_on='feature.name',
    fill_color='RdPu',
    fill_opacity=0.8,
    legend_name='The average cost of a cup of cappuccino in Moscow districts',
).add_to(m)

# displaying the map
m

Make this Notebook Trusted to load map: File -> Trust Notebook

The most expensive cappuccino in the coffee shops of the Central and Southwestern administrative districts — an average of 187 rubles per cup.

Earlier, we also studied the distribution of chain and non-chain places by category and know that in Moscow the number of chain and non-chain coffee shops is approximately equal, unlike cafes, restaurants and bars, where there are significantly more non-chain establishments.

The average rating of chain and non-chain coffee shops in Moscow districts.

districts_coffee_shops_chain_rating = data.query('category == "coffee shop"').groupby(['district_eng', 'chain']).agg(rating=('rating', 'mean')).reset_index()
districts_coffee_shops_chain_rating['rating'] = districts_coffee_shops_chain_rating['rating'].round(2)
districts_coffee_shops_chain_rating['district_eng'] = districts_coffee_shops_chain_rating['district_eng'].apply(lambda x: x.split(' ')[0].strip())
districts_coffee_shops_chain_rating

	district_eng	chain	rating
0	Central	0	4.40
1	Central	1	4.28
2	Eastern	0	4.35
3	Eastern	1	4.21
4	North-Eastern	0	4.28
5	North-Eastern	1	4.15
6	North-Western	0	4.49
7	North-Western	1	4.19
8	Northern	0	4.38
9	Northern	1	4.20
10	South-Eastern	0	4.32
11	South-Eastern	1	4.03
12	South-Western	0	4.32
13	South-Western	1	4.25
14	Southern	0	4.32
15	Southern	1	4.15
16	Western	0	4.24
17	Western	1	4.17

plt.figure(figsize=(14,7))
plt.title('Distribution of average ratings of chain and non-chain coffee shops by districts of Moscow',fontsize=15)

g = sns.barplot(data=districts_coffee_shops_chain_rating, x='district_eng', y='rating', hue='chain')
for p in g.patches:
    g.annotate('{:.2f}'.format(p.get_height()), xy=(p.get_x() + p.get_width() / 2, p.get_height()), ha = 'center', va = 'center', xytext = (0, 10), textcoords = 'offset points')
plt.xlabel('Administrative districts of Moscow',fontsize=12)
plt.ylabel('Average rating',fontsize=12)
g.set_ylim(4, 4.55);

In each administrative district of Moscow, the rating of non-chain coffee shops is higher than the chain ones.

Conclusion

• There are 1,413 coffee shops in total. As we know from the study of the distribution of foodservice places by districts of Moscow, in the Southern Administrative District, the number of places of which is comparable to the Northern and North-Eastern District, there are relatively fewer coffee shops. From the point of view of the share of coffee shops among all the places of the district, one can consider the Southern Administrative District of Moscow to open a coffee shop.

• There is only one round-the-clock coffee shop in the South-Eastern and Southern district of Moscow. The largest number of 24—hour coffee shops is in the city center (26).

• In Southern Administrative District, the average rating of coffee shops is slightly lower than in Moscow as a whole.

• The cost of a cup of cappuccino in the Southern Administrative District is 165 rubles. We recommend focusing on the cost not lower than this.

• In each administrative district of Moscow, the rating of non-chain coffee shops is higher than chain ones. In the Southern Administrative District, this difference is especially noticeable: the rating of non-chain places is on average 4.32 versus 4.03 for chain places.

We recommend opening a non-chain 24-hour coffee shop in the Southern Administrative District of Moscow with the cost of a cup of cappuccino from 165 rubles.

General conclusion

1. We have studied a dataset with information about 8,406 foodservice places in Moscow.

2. Prepared the dataset for work

The missing values were processed:

• the price column was filled in for chain places by analogy with other chain places located in the same administrative district;
• the middle_avg_bill column was filled with the median value for places of the same category with the same price level and located in the same administrative district;
• column middle_coffee_cup was filled in for chain coffee shops by analogy with other chain establishments that are located in the same administrative district.

We added a column street with street names to the dataset , a column is_24/7 with the designation that the places is open daily and around the clock.

1. We studied the data and came to the following conclusions about catering establishments in Moscow:

• The dataset has the most cafes and restaurants. Restaurants, bars/pubs and coffee shops are in the lead in terms of the number of seats. There are 63.2% of non—chain places in the dataset, 36.8% of chain places.

• Domino's Pizza and Dodo Pizza pizza chains are leading among the popular chains. The top 15 chains do not include pubs/bars, fast food places, canteens.

• Most of the places from the top 15 chains are in the Central Administrative District of Moscow (11). Least of all — one — in the East. The Central Administrative District of Moscow is also the leader in the total number of places — 2,237.

• The average rating of all categories of places exceeds 4 points. Bars/pubs have the maximum rating (4.39). The highest proportion of places with a rating below 4 points are fast food places (27%) and cafes (23%).

• Mira Avenue is the leader in terms of the number of foodservice places in Moscow, 184 places are located on this street. There are 458 streets where there is only one foodservice place in Moscow. Most often the only place will be a cafe.

• The median average bill in the Central Administrative District of Moscow (950 rubles) is significantly higher than in other districts. The cheapest places are located in the South-Eastern district (the average bill is 425 rubles.)

• The highest percentage of round-the-clock places in the fast food category, the lowest in canteens.

1. We studied the coffee shops in Moscow and came to the following conclusions:

• There are 1,413 coffee shops in total. In the Southern Administrative District, the number of places of which is comparable to the Northern and North-Eastern districts, there are relatively fewer coffee shops.

• There is only one round-the-clock coffee shop in the South-Eastern and Southern districts of Moscow. The largest number of 24—hour coffee shops is in the city center (26).

• In Southern Administrative District, the average rating of coffee shops is slightly lower than in Moscow as a whole.

• The cost of a cup of cappuccino in the Southern Administrative District is 165 rubles .

• In each administrative district of Moscow, the rating of non-chain coffee shops is higher than chain ones. In the Southern Administrative District, this difference is especially noticeable: the rating of non-chain places is on average 4.32 versus 4.03 for chain places.

According to the results of a detailed study, it was recommended to open a non-chain 24-hour coffee shop in the Southern Administrative District of Moscow with the cost of a cup of cappuccino from 165 rubles.